Strategies in Cochlear Nerve Regeneration, Guidance and Protection: Prospects for Future Cochlear Implants

(1)

UNIVERSITATISACTA

Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine 1193

Strategies in Cochlear Nerve Regeneration, Guidance and Protection

Prospects for Future Cochlear Implants

FREDRIK EDIN

(2)

Dissertation presented at Uppsala University to be publicly examined in Skoogsalen, Akademiska Sjukhuset, Ingång 78/79, Uppsala, Thursday, 28 April 2016 at 09:00 for the degree of Doctor of Philosophy. The examination will be conducted in English. Faculty examiner: Docent Maoli Duan (Karolinska Institutet, Department of Clinical Science, Intervention and Technology).

Abstract

Edin, F. 2016. Strategies in Cochlear Nerve Regeneration, Guidance and Protection. Prospects for Future Cochlear Implants. Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine 1193. 56 pp. Uppsala: Acta Universitatis Upsaliensis.

ISBN 978-91-554-9503-9.

Today, it is possible to restore hearing in congenitally deaf children and severely hearing- impaired adults through cochlear implants (CIs). A CI consists of an external sound processor that provides acoustically induced signals to an internal receiver. The receiver feeds information to an electrode array inserted into the fluid-filled cochlea, where it provides direct electrical stimulation to the auditory nerve. Despite its great success, there is still room for improvement, so as to provide the patient with better frequency resolution, pitch information for music and speech perception and overall improved quality of sound.

A better stimulation mode for the auditory nerves by increasing the number of stimulation points is believed to be a part of the solution. Current technology depends on strong electrical pulses to overcome the anatomical gap between neurons and the CI. The spreading of currents limits the number of stimulation points due to signal overlap and crosstalk.

Closing the anatomical gap between spiral ganglion neurons and the CI could lower the stimulation thresholds, reduce current spread, and generate a more discrete stimulation of individual neurons. This strategy may depend on the regenerative capacity of auditory neurons, and the ability to attract and guide them to the electrode and bridge the gap.

Here, we investigated the potential of cultured human and murine neurons from primary inner ear tissue and human neural progenitor cells to traverse this gap through an extracellular matrix gel.Furthermore, nanoparticles were used as reservoirs for neural attractants and applied to CI electrode surfaces. The nanoparticles retained growth factors, and inner ear neurons showed affinity for the reservoirs in vitro.

The potential to obtain a more ordered neural growth on a patterned, electrically conducting nanocrystalline diamond surface was also examined. Successful growth of auditory neurons that attached and grew on the patterned substrate was observed.

By combining the patterned diamond surfaces with nanoparticle-based reservoirs and nerve- stimulating gels, a novel, high resolution CI may be created. This strategy could potentially enable the use of hundreds of stimulation points compared to the 12 – 22 used today. This could greatly improve the hearing sensation for many CI recipients.

Keywords: Human vestibular nerve, Scarpa's ganglion, Stem cells, Nanoparticles, Nanocrystalline diamond

Fredrik Edin,

urn:nbn:se:uu:diva-276336 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-276336)

(3)

Till Morfar

(4)

Front cover: Double immunofluorescence staining of a human vestibular ganglion shows large neural cell bodies and neural extensions staining positive for Tuj1 (red) and TrkB (green). Nuclei are stained with DAPI (blue). Picture adapted from Paper II.

(5)

List of Papers

This thesis is based on the following papers, which are referred to in the text by their Roman numerals.

I Edin F, Liu W, Boström M, Magnusson PU, Rask-Andersen H. Differentiation of human neural progenitor cell-derived spiral ganglion-like neurons: a time-lapse video study. Acta Oto- laryngol. 2014 May;134(5):441-7.

II Edin F, Liu W, Li H, Atturo F, Magnusson PU, Rask- Andersen H. 3-D gel Culture and Time Lapse Video Micros- copy of the Human Vestibular Nerve. Acta Otolaryngol. 2014 Dec;134(12):1211-8.

III Li H, Edin F, Hayashi H, Gudjonsson O, Engqvist H, Rask- Andersen H and Xia W. Guided Growth of Auditory Neurons:

Bioactive Particles Towards Gapless Neural - Electrode Inter- face. Submitted Manuscript.

IV Cai Y, Edin F, Jin Z, Alexsson A, Gudjonsson O, Liu W, Rask-Andersen H, Karlsson M, and Li H. Strategy towards In- dependent Electrical Stimulation from Cochlear Implants:

Guided Auditory Neuron Growth on Topographically Modi- fied Nanocrystalline Diamond. Acta Biomater. 2016 Feb;31:211-220.

Reprints were made with permission from the respective publishers.

(6)

Papers not included in thesis

I Brown TD, Edin F, Detta N, Skelton AD, Hutmacher DW, Dalton PD. Melt electrospinning of poly(ε-caprolactone) scaf- folds: phenomenological observations associated with collec- tion and direct writing. Mater Sci Eng C Mater Biol Appl.

2014 Dec;45:698-708.

II Nordling S, Hong J, Fromell K, Edin F, Brännström J, Lars- son R, Nilsson B, Magnusson PU. Vascular repair utilising immobilised heparin conjugate for protection against early ac- tivation of inflammation and coagulation. Thromb Haemost.

2015 Jun;113(6):1312-22.

III Liu W, Edin F, Atturo F, Rieger G, Lowenheim H, et al. The Pre- and Post-somatic segments of the Human Type I Spiral Ganglion Neurons - Structural and Functional Considerations Related to Cochlear Implantation. Neuroscience. 2015 Jan 22;284:470-82.

IV Liu W, Edin F, Blom H, Magnusson PU, Schrott-Fischer A, Glueckert R, Santi P, Li H, Laurell G and Rask-Andersen G.

Super-Resolution Structured Illumination Fluorescence Mi- croscopy of the Lateral Wall of the Cochlea –the Connex- in26/30 Proteins are Separately Expressed in Man. Cell Tissue Res 2016. Early online.

V Natan M, Edin F, Perkas N, Yacobi G, Perelshtein I, Segal E, Homsy A, Laux E, Keppner H, Rask-Andersen H, Gedanken A and Banin E. Two Are Better than One: Combining ZnO and MgF2 Nanoparticles Reduces Streptococcus pneumoniae Biofilm Formation on Cochlear Implants. Accepted in Adv.

Func. Mat. 2016.

(7)

VI Liu W, Edin F, Brännström J, Glueckert R, Santi P, Salven- moser W, Laurell G, Blom H, Schrott-Fischer A and Rask- Andersen H. The Epithelial Gap Junction Network in the Hu- man Cochlea. An Ultrastructural, Laser Confocal and Super- resolution Structured Illumination Microscopy Study. Submit- ted Manuscript.

(8)

(9)

Abbreviations

ABI Auditory brainstem implant

ABR Auditory brainstem response

BDNF Brain-derived neurotrophic factor

bFGF Basic fibroblast growth factor

BM Basilar membrane

CI Cochlear implant

CNT Carbon nanotubes

CPHS Calcium phosphate hollow nano-

spheres

DMEM Dulbecco’s modified eagles medium

EGF Epidermal growth factor

ECM Extracellular matrix

ES Embryonic stem (cells)

FGF Fibroblast growth factor

GDNF Glial cell line-derived neurotrophic

factor

HEI House Ear Institute

hNPC Human neural progenitor cell

IHC Inner hair cell

IPS Induced pluripotent stem cell

LIF Leukemia inhibitory factor

NCD Nanocrystalline diamond

NT-3 Neurotrophin-3

OC Organ of Corti

OHC Outer hair cell

PFA Paraformaldehyde PI Platinum-iridium

SG Spiral ganglion

SNHL Sensory neural hearing loss

TLVM Time-lapse video microscopy

TM Tectorial membrane

TrkB Tropomyosin related kinase B,

BDNF receptor (a.k.a. NTRK2)

Tuj1 Neuron-specific class III beta-tubulin

VG Vestibular ganglion

(12)

(13)

Introduction

The inner ear, hearing, and hearing loss

The human inner ear is one of the most intriguing structures of the human body. It relies on different highly specialized cell types set in a precise order to execute its functions. Landmark discoveries about the ear were made by Antonio Scarpa (1752 – 1832) and Alfonso Corti (1822 – 1876), who also lent their names the structures they discovered. At the end of the 19^th century, the anatomy of the inner ear in humans and several animals was described in remarkable detail by the Swedish anatomist Gustaf Retzius (Retzius, 1884).

The inner ear consists of two parts, with separate functions, that both de- velop from the otocyst (Fekete, 1999). The cochlea is responsible for hearing, while the vestibular organ provides the senses of balance and motion.

Innervation of the two organs is separated. The spiral ganglion (SG) innervates the cochlea and is connected to the brainstem via the auditory nerve.

The vestibular nerve relays information from the vestibular ganglion (VG, or Scarpa’s ganglion) which innervates the balance organ. Together, the auditory and vestibular nerves form the VIII^thcranial nerve.

Sound consists of vibrations. When sound reaches the external ear, which consists of the pinna and external ear canal, it generates minute vibrations in the tympanic membrane. The tympanic membrane is connected to the ossicular chain, consisting of the maleus, incus and stapes. This is a highly tuned system, capable of amplifying sound in a frequency-dependent man- ner, both in the external ear canal and the ossicles. The ossicular chain trans- fers mechanical vibrations via the middle ear to the inner ear. The stapes emits vibrations to the oval window of the cochlea, which causes pressure changes in the cochlear fluids. Mechanosensory receptors, called inner hair cells (IHC), inside the cochlea then transform the mechanical vibrations into electrical impulses, transferred to the brainstem via the auditory nerve.

When the pressure changes reach the cochlear fluids, they give rise to a traveling wave in the basilar membrane (BM), which separates the scala tympani from the scala media (Von Békésy & Wever, 1960). The BM con- tributes to the cochlear frequency resolution as, from base to apex, its width linearly increases four times whilst the thickness decreases six times (Khanna & Leonard, 1982; Patuzzi et al., 1982; Liu et al., 2015).

(14)

The organ of Corti (OC) sits on the BM, beneath the tectorial membrane (TM) and contains the outer and inner hair cells (Figure 1). The BM move- ments result in shearing forces between the stereociliary bundles of the outer hair cells (OHCs) and the TM. The highly specialized OHCs can act as local amplifiers and their mechanical response increases not only sound sensitivity but also frequency resolution. The TM is also believed to contribute to coch- lear tuning, but the exact mechanisms remain to be elucidated (Ghaffari et al., 2007; Hayashi et al., 2015). The 3,400 IHC then convert the vibrations into electrical signals. The IHC are connected to approximately 30,000 Type I neurons that mediate afferent signals to the brain via the VIII^th cranial nerve (Otte et al., 1978).

Consequently, the BM and OHCs may act as primary and secondary fil- ters partly responsible for the highly precise frequency resolution in the mammalian cochlea. Together they shape a finely tuned, tonotopically ar- ranged inner ear map (Greenwood, 1961). The highest frequency sounds are filtered near the round window, whereas the lowest frequencies are filtered near the apex (Liberman, 1982).

According to the World Health Organization (WHO, 2013), more than 5

% of the world’s population suffers from disabling hearing loss. This is often caused by infections, ototoxic drugs, noise exposure, or age-related hearing loss (also known as presbyacusis) (Niihori et al., 2015).

Fragmented OHC loss as seen in presbyacusis reduces hearing sensitivity and capacity in noisy environments. It can often be amended through sound amplification using hearing aids. When IHCs are lost, on the other hand, the cochlea progressively loses its afferents, resulting in gradual deafness. In severe cases of inner ear associated hearing loss, also known as sensory- neural hearing loss (SNHL), the most common treatment is cochlear im- plants (CI) (Shepherd et al., 1993).

Figure 1. Cross section showing the sensory organ (Organ of Corti) and the three fluid filled spaces of the cochlea. The scala vestibuli is connected to the oval window where the stapes attaches. Cochlear implants are inserted into the scala tympani.

Picture adapted from work by Oarih Ropshkow, available according to CC BY-SA.

(15)

Cochlear implants

The first electrically evoked sensation of hearing was reported by Ales- sandro Volta in the late 18^th century. He connected a series of batteries to two rods which were inserted into fluid-filled external ear canals (Volta, 1800). Apart from being extremely unpleasant, the effect was described as the sound of boiling thick soup.

Groundbreaking results in the 1930s showed that, at certain frequencies, the original sound could be reconstructed from signals recorded from the auditory nerve of a cat while providing sound stimulus. This was the first evidence of a correlation between perceived sound and the signaling to the brain (Wever & Bray, 1930).

In 1957, the first human trial using direct electrical stimulation of the auditory nerve was performed by French clinicians. A sense of sound was evoked by electrically stimulating part of the VIII^th cranial nerve in a patient who had lost both inner ears during surgery (Djourno & Eyries, 1957). Alt- hough the implant was later removed, the patient received some pitch perception.

Current CI technology stems from the 1970s and the first single channel electrode that was inserted into the cochlea of a deaf patient by the pioneer William House (Mudry & Mills, 2013). He worked at the House Ear Insti- tute (HEI) in Los Angeles, US. This institute was for a long time a leader in auditory implant innovation and development, being one of the first CI cen- ters, and the first center to implant a child with a CI in the US. Several research laboratories worked in parallel, and by the end of the 1970s the first multichannel CI was implanted in France (Chouard, 2015), followed by their commercial launch in the 1980s (Wilson, 2013).

CI electrodes are inserted into the scala tympani, either through a cochle- ostomy or the less traumatic round window approach (Adunka et al., 2004b).

A CI acts through a direct electrical stimulation of the auditory nerve, by- passing dysfunctional IHCs (Michelson & Schindler, 1983). The exact site of electrical stimulation of the auditory nerve remains elusive and it is cur- rently unknown if the stimulation activates voltage-gated ion channels on the SG soma, the hillock region, or at the Ranvier´s node of the peripheral or proximal afferents.

(16)

Figure 2. Cochlear implants. The external processor and transmitter sit behind the external ear and connect wirelessly to the subcutaneous receiver. The receiver is connected to the electrodes placed inside the cochlea. Picture by MED-EL.

Sound is collected by a microphone situated on the external processor. The acoustic signal is transformed into several band-passed electrical signals that are wirelessly transmitted through the skin to a subcutaneous receiver. The receiver decodes the signals into electrical pulses that stimulate the SG through multiple stimulation points on the electrode placed inside the coch- lea (Figure 2 and Figure 3 C). Speech can be perceived from temporal cues, but the fine structure and spectral features are difficult to convey through the implant (Hochmair et al., 2015).

The implants are generally made from flexible silicone and wrap around the helical, central structure of the cochlea, also known as the modiolus.

Typically a CI has 12 – 22 electrodes (stimulation points) composed of an inert platinum-iridium (PI) alloy. Electrical stimulation of a certain region of the modiolus will produce a hearing sensation in a particular frequency region as neural responses in the SG are place-mapped, roughly following the aforementioned tonotopic map (Stakhovskaya et al., 2007). This means that a pulse near the round window is perceived as a high-pitched sound whereas an apical pulse will provide a lower frequency sound. However, the electrical pulse characteristics, such as pulse rate and amplitude, influence the psy- chophysical outcome and perception which is taken advantage of in modern CI processors.

Through hearing preservation, surgery patients with residual hearing can receive so-called hybrid hearing or electro-acoustic stimulation, which relies on both acoustic and electrical stimulation (Kiefer et al., 2004). The patient can perceive low frequency sounds through their natural hearing while high frequency sounds are mediated through the CI electrical stimulation. This strategy often gives the patient a softer and more natural hearing sensation, including better speech perception in noisy surroundings (Erixon et al., 2012).

(17)

In recent years, development of more atraumatic surgical techniques that allow for structural preservation has been considered essential. As it is im- possible to foresee how future SNHL therapies may act, any structural damage caused during implantation may prevent a patient from receiving novel therapies when the previous implant has to be replaced. Life expectancy in many industrialized countries today is around 80 years (U.N., 2012). This means that implanted children will, most likely, go through at least one re- implantation during their lifetime (Hochmair et al., 2015).

Thus far, more than 450,000 people have received a CI and the number of implant recipients is increasing exponentially, suggesting that more than one million people will receive a CI by 2020. Thanks to the cochlear implant technology, deaf children can today be taught in regular schools, and adults can access job opportunities and maintain connections with their families, friends, and society.

Auditory brainstem implants

The use of CI is limited to patients with an innervated cochlea. If the auditory nerve is lacking or dysfunctional, an auditory brainstem implant (ABI) electrode can be placed directly on the brain stem and provide hearing sensa- tions through direct electrical stimulation of the cochlear nucleus (Vincent, 2012). A patient group that might receive an ABI is those suffering from bilateral acoustic tumors as they often lose both of their auditory nerves during surgery. An ABI can then be implanted to provide some sense of hearing (Siegbahn et al., 2014).

ABI results are generally poorer in terms of speech perception and quality of hearing quality compared to a CI, most likely due to anatomical variations in frequency mapping (Lundin et al., 2015).

Approaches to improve CI outcome

Electrode-based strategies

Various strategies have been developed to improve CI performance. For instance, it would be desirable to increase the number of stimulation points to provide a more discrete stimulation in each frequency region, as this is thought to be beneficial for speech perception and the ability to enjoy music (Wilson, 2013). This has proven difficult since tissue in the anatomical gap between CI and SG acts as resistance while the perilymph is highly conduc- tive, causing wide current spread in the cochlea. By limiting or eliminating the gap, the electrodes could employ less powerful pulses to evoke a SG

(18)

neuron response (Shepherd et al., 1993) which has also been demonstrated in vitro (Hahnewald et al., 2015). This in turn would limit current spread and allow the stimulation points to be positioned more closely without risking stimulation overlap, allowing for higher resolution CI (O'Leary et al., 2009;

Wilson, 2013) as well as lower power consumption.

The modiolus hugging electrode, also known as the perimodiolar electrode, is pre-shaped to wrap around the modiolus as it is inserted. These electrodes are already in use and were found to improve speech understanding in clinical studies (Bacciu et al., 2004). They have also exhibited decreased power consumption and stimulation thresholds (Wackym et al., 2004) com- pared to straight electrodes. One downside of this electrode type is that the increased size often necessitates insertion through a cochleostomy (Doshi et al., 2015) which could cause unnecessary trauma (Adunka et al., 2004a), particularly in patients with residual hearing.

To further decrease the distance, attempts have been made to combine perimodiolar implants with micro-needles to penetrate the wall of the modio- lus, providing intra-modiolar stimulation (Badi et al., 2002). Other layouts for intra-modiolar electrodes have been designed with 194 stimulation points (Volckaerts et al., 2007). Neither of the two strategies has reached the clinic.

Signal focusing can also be achieved by grounding, or providing negative currents through electrodes neighboring the stimulating electrode. Studies employing such strategies have shown that neural responses are more limited and focused, which indicates a more selective stimulation (George et al., 2014). Focusing of the signal comes at a cost, since stimulation thresholds increase compared to monopolar stimulation, causing higher power con- sumption (George et al., 2015). As the strategy has already been tested clini- cally and regular CI electrodes can be used, this may very well be the next step in electrode improvement (van den Honert & Kelsall, 2007).

Cell- and gene- based therapies against hearing loss

Stem cells

For many years, embryonic stem (ES) cells have been the main focus of stem cell research. These cells are capable of proliferating indefinitely and are pluripotent, which means that they are capable of differentiating into cells of all three germ layers (Thomson et al., 1998). Stem cell populations have been found also in adult tissue; these are so-called progenitor cells, which slowly divide and provide new cells, e.g. for skin or intestinal villi.

Innate, adult stem cell populations have also been found in the nervous sys- tem and even the adult cochlea (Rask-Andersen et al., 2005). These progeni- tor cells generally have a more limited potential for differentiation.

(19)

In recent years, the induced pluripotent stem (IPS) cells have come into focus. Any somatic cell can, in theory, be transformed into an IPS cell by transfection with the genes Oct3/4, Sox2 and Nanog (Takahashi &

Yamanaka, 2006). This technique was awarded the Nobel Prize in 2012 and is considered to have great promise in future therapies, as the patient’s own cells can be reprogrammed to supplement damaged tissue.

Both IPS and ES cells are considered as important in vitro supplements to animal models, and may have potential in the treatment of neurodegenerative diseases such as Parkinson’s and Alzheimer’s.

Restoring the auditory nerve

The performance of a CI may depend on the status of the auditory nerve, but a clear correlation between implantation outcome and the remaining number of neurons does not seem to exist (Fayad & Linthicum, 2006). It has been suggested that as little as 10 % of the SG population is sufficient for a CI to function and give a satisfactory outcome (Linthicum et al., 1991).

Supplementation of neurons to the auditory nerve, either by stem cells or nerve grafts, has been considered as a strategy to improve CI outcome. This treatment could benefit cases where the auditory nerve has been severely degenerated by disease or lost during surgery. Such treatments could also be a prerequisite for future generations of CI, which may require a larger neural population to function (Olivius et al., 2003; Olivius et al., 2004; Palmgren et al., 2011; Chen et al., 2012; Palmgren et al., 2012). The success of the trans- plant would depend on the successful integration of the newly formed cells into the auditory nerve and the formation of functional connections to the brainstem.

Animal studies have shown that transplanted neurons can re-innervate IHCs in a denervated OC (Martinez-Monedero et al., 2006; Martinez- Monedero & Edge, 2007; Chen et al., 2012). Improved hearing thresholds in deafened animals after stem cell injections have also been reported, suggest- ing that functional connections to the brainstem can be achieved (Chen et al., 2012).

The anatomy of the cochlea makes intra-cochlear administration of stem cells difficult without causing irreversible structural damage. By injecting the stem cells into the auditory nerve, these risks could be reduced, since the cochlea would not be opened (Corrales et al., 2006; Palmgren et al., 2012).

The risk of tumor formations, both benign and malignant, must be taken into account when evaluating the potential benefits of stem cell transplanta- tion (Amariglio et al., 2009).

(20)

Hair cell regeneration

Hair cell regeneration would be a giant leap in hearing restoration and make the CI obsolete. Unfortunately, the complex and highly ordered structures would easily be damaged by a stem cell injection, and it would be difficult to target stem cells towards the OC. Regenerating hair cells and integrating them into the OC would most likely require differentiation of the adjacent supporting cells, normally surrounding the hair cells (Mizutari et al., 2013;

Bramhall et al., 2014). As the supporting cells degenerate after hair cell loss, leaving a flat epithelium, treatments stimulating differentiation would need to be administered before the OC degenerates for this to be a realistic option (Izumikawa et al., 2008).

The stem cell source is also an important issue. Unless IPS or innate stem cells isolated from the tissue of the patient are used, chemical immunosup- pressive treatments would be needed to avoid rejection (Nussbaum et al., 2007). Presumably, we will see further improvements of the CI technology before stem cell-based therapies will become a reality.

(21)

The NanoCI project

The work in this thesis was largely funded and conceptualized within the project “Nanotechnology based cochlear implant with gapless interface to auditory neurons”, with the acronym “NanoCI”, within the 7^th framework program of the EU. The project started September 2012 and ended August 2015 and aimed at closing the anatomical gap between CI and SG. The idea of the project was based on results from the aforementioned animal studies, which showed that SG can re-sprout peripheral extensions towards the scala tympani after hair cell loss (Glueckert et al., 2008). To achieve this, factors providing stimulation and guidance to neurons were developed and produced within the consortium, as well as matrices to replace the perilymph of the scala tympani (Figure 3). Uppsala University, along with collaborators at the Universities of Tübingen, Germany and Bern, Switzerland, were involved in the in vitro testing of these factors. The group in Tübingen also performed in vivo studies based on the most promising in vitro results at the end of the project. All work is summarized into reports available to the public through the NanoCI consortium home page (www.nanoci.org).

Figure 3. The NanoCI project. Images show scanning electron microscopy images of half of a human cochlea cut through the modiolus to reveal internal structures. A) Showing approximate innervation (yellow) of SG in one turn of an ear with normal hearing. B) After hair cell loss, peripheral extensions degenerate leaving monopolar SG neurons. C) A regular CI placed in the scala tympani to stimulate the monopolar SG neurons. D) NanoCI approach, with a gel replacing the perilymph, and an implant with neurotrophin reservoirs stimulating regeneration of the peripheral extensions of the SG and guiding them towards a modified implant. Scale bars 1 mm.

Adapted from Rask-Andersen et al. (2012).

(22)

Auditory nerve regeneration

The presence of stem cells in rodents and even in the adult human cochlea has been previously described, indicating that there is a certain innate regen- erative capacity in the mammalian inner ear (Rask-Andersen et al., 2005;

Oshima et al., 2009). Murine SG and vestibular ganglion (VG) can also re- sprout neural extensions in a Matrigel-like matrix in vitro (Gaboyard et al., 2005), and paper II shows the potential for human VG do the same.

Human SG neurons can remain electrically excitable in a monopolar form for decades after IHC loss thus enabling the use of CI (Linthicum & Fayad, 2009). As they are still viable, they might be susceptible to neurotrophic stimulation. In contrast, some mammals show a dramatic decrease in SG neuron numbers that can be observed within months after deafening (Dodson

& Mohuiddin, 2000). In deafened guinea pigs, only around 5 % of the SG neuron population remains after long term deafness (Dodson, 1997).

The discrepancy between SG survival in humans and animals may not be entirely physiological. It could to some extent be explained by the harsh deafening conditions used in animal experiments (Kong et al., 2010).

Various approaches have been used to deliver neurotrophins into the cochlea in animal studies. The aim of the studies is often to maintain the SG population after deafening, as well as to encourage re-sprouting of peripheral ex- tensions (Glueckert et al., 2008; Wise et al., 2010; Pinyon et al., 2014).

One approach employs chronic infusion of growth factors through a cath- eter in the round window. Using this method, brain-derived neurotrophic factor (BDNF) and fibroblast growth factor (FGF) was delivered to the cochlea of guinea pigs, which stimulated peripheral extensions to invade the scala tympani (Glueckert et al., 2008). Using an external supply permits long-term administration at high concentrations, but the risk of infections may out- weigh the advantages.

Another approach is to deliver an intra-cochlear neurotrophin supply. One method that has been suggested is to place neurotrophin-producing cells on the surface of the CI (Rejali et al., 2007). Techniques to genetically modify cochlear cells in situ into producing the desired neurotrophins have also been considered (Wise et al., 2010; Pinyon et al., 2014). Receiving clinical ap- proval for these approaches could prove difficult, due to fears of tumor formation and immunological reactions. Providing factors for an extended time period merely to maintain SG neuron populations may also prove unnecessary in humans, as previously mentioned (Linthicum & Fayad, 2009).

If the therapies aim at re-sprouting and guiding dendrites towards a CI, the neurotrophin reservoirs may only need to last for a limited amount of time. Adding biodegradable reservoirs into the scala tympani (Wang et al., 2014) or directly onto the CI could provide a sufficiently long release of the guidance cue. To further assist the bridging of the gap between modiolus and

(23)

CI, the distance could first be decreased using a perimodiolar implant. A gel could be used to replace the perilymph and facilitate neural outgrowth (Xie et al., 2013). Conditioning the gel with growth factors would provide an additional neurotrophin source (Hutten et al., 2014).

A potential carrier of growth factors that could be applied directly to the CI is calcium phosphate hollow nanospheres (CPHS). Calcium phosphates are the major constituent of bone and teeth and are both biocompatible and biodegradable. They can also bind proteins and drugs (Ginebra et al., 2012).

By introducing pores and making the structure hollow, the surface area for binding growth factors is increased; this also makes the nanoparticles easier to degrade.

Electrode surface modifications

The interest in using nano-patterned diamond substrates made from nanocrystalline diamond (NCD) has increased greatly in the past years. NCD surfaces have been found to be biocompatible, durable (Amaral et al., 2008), have anti-bacterial properties (Jakubowski et al., 2004), and promote for- mation of neural networks. Neurons cultured on the NCD surfaces can also form electrically active networks (Thalhammer et al., 2010).

The field of artificial retinal replacements is a research area where the need for an increased number of stimulation points was identified early. It has been estimated that such an implant will require at least 600 functional pixels to distinguish between faces and for reading (Bendali et al., 2015).

This is a vast number considering the 12 – 22 stimulation points of the current CI. To reach 600 stimulation points the retinal implants and CI will have to overcome similar technical challenges, albeit with a more accessible neural population in the eye. 64-channel, flexible implant prototypes have al- ready been produced with NCD and in vivo studies have shown that they can be implanted to the retina with minimal glial scarring (Bergonzo et al., 2011).

Diamond surfaces structured with wells to limit current spread stimulated retinal glia and bipolar cells to integrate with the implant (Djilas et al., 2011). This approach could potentially be suitable for an ABI rather than a CI as it is positioned in direct contact with the neural tissue in the brainstem.

By replacing the PI electrodes of the CI with nanostructured NCD, a gapless CI could potentially have thousands of stimulation points to provide an improved frequency discrimination.

(24)

Aims

The overall aim of this thesis is to study the regenerative capacity of the human VIII^th cranial nerve and the implications for development of a novel cochlear implant.

Paper I

To evaluate the human neural progenitor cell (hNPC) line ENstem-A as a potential source for SG-like cells.

Paper II

To establish a 3-D culture model for human vestibular ganglions collected during vestibular schwannoma surgery.

Paper III

To investigate the efficacy of a nanoparticle-based slow release system and establish a model for neural guidance in vitro.

Paper IV

To investigate the potential of patterned diamond substrates to act as a new surface for an interfaceable cochlear implant.

(25)

Material & Methods

Cell culture media

All cultures were maintained using the following media compositions:

Table 1. Neurobasal-based culture media (Papers I – IV).

”Neurobasal media” Concentration Supplier

Neurobasal – A Gibco

B27 supplement 1:50 Gibco

L-glutamin 1 mM Gibco

Gentamicine 0.04 % Gibco

Concentrations of growth factors added in ”Expansion media”

bFGF 20 ng/mL Millipore

LIF 10 ng/mL Millipore

Concentrations of growth factors added in “Complete media”

BDNF 20 ng/mL R&D Systems

GDNF 20 ng/mL R&D Systems

NT-3 20 ng/mL R&D Systems

Table 2. DMEM/F12 –based culture media (Papers II, IV).

”DMEM/F12 media” Concentration Supplier

DMEM/F12 Gibco

B27 supplement 1:50 Gibco

L-glutamin 1 mM Gibco

N2 supplement 1:10 Gibco

Gemtamicine 0.04 % Gibco

EGF 20 ng/mL Austral Biologicals

bFGF 10 ng/mL Millipore

Heparan sulfate 50 µg/mL Sigma-Aldrich

(26)

Stem cell culture (Paper I, II, IV)

Two commercially available neural progenitor lines were used.

In paper I the non-immortalized, h9-derived, hNPCs called ENstem-A (Millipore) were used. They were proliferated in “Expansion media” containing leukemia inhibitory factor (LIF) and basic fibroblast growth factor (bFGF) which promote expansion and limits differentiation.

They were typically passaged 1:3 Accutase (Life Technologies) or Try- plEx (Life Technologies) as Accutase gradually became ineffective at disso- ciating cells.

For experiments, cells were counted in a TC20 Automated cell counter (Bio-Rad) and typically seeded at a density of approximately 3,000 cells/cm².

For differentiation “Complete media” was used. The trio of growth factors, BDNF, GDNF and neurotrophin 3 (NT-3) at a 20 ng/mL concentration has previously been used to differentiate and maintain SG neurons in primary culture, where they were found to be most effective when administered together (Rask-Andersen et al., 2005; Boström et al., 2010).

In paper IV the immortalized ReNcell line (Millipore) was used. These cells are also neural progenitors but are, in our hands, easier to culture than ENStem-A.

These cells were only used in a non-differentiated state in DMEM/F12 media. Passaging was performed using a suspension of 0.1 wt% Trypsin and 0.002 wt% EDTA (Life technologies). Cells were counted manually using in a Bürker chamber.

Primary cultures (Paper II – IV)

Human tissues were harvested in accordance with ethical permits (no.

99398, 22/9 1999, cont., 2003 and Dnr. 2013/190), with patient consent, and in accordance with the Helsinki declaration, during surgery to remove vestibular schwannomas or petroclival meningiomas using a translabyrinthine approach. In the operating theater, the tissue was immediately placed in cooled Leibovitz’s L15 media. It was then dissected under dissection microscope into pieces suitable for culture, between 0.5 – 1 mm in diameter. After washing in PBS, explants were embedded in Matrigel-gel, containing 1:1 of Matrigel (8 mg/mL) and DMEM/F12 media further supplemented with 50 ng/mL of IGF (Sigma-Aldrich). The gel was allowed to set for 30 min at 37

°C, 5 % CO2, after which it was covered with the supplemented DMEM/F12 media. After 24 hours the media was replaced with “Complete media” to stimulate neural outgrowth.

(27)

As the human material is rare, SG tissue from guinea pigs was also used, in accordance with ethical permits (C98/12). SG tissue from P9 neonatal mice was also used and was harvested from residual tissue in accordance with supplemented ethical approvals (C346/11 and C1/15). After dissecting out the cochlea and isolating the SG, this tissue was treated in the same way as the human material.

Time-lapse video microscopy (Paper I – III)

Time-lapse video microscopy (TLVM) material was mainly captured using a Nikon TE2000-E fitted with a CO2 incubator and temperature control as well as a perfect focus system. Pictures were taken every 1 – 3 minutes. Playback rate was typically 14 – 15 fps.

Immunofluorescence (Paper I – IV)

Cells and tissues were fixated for 20 – 30 minutes in 4 % fresh paraformaldehyde (PFA) at room temperature. After rinsing, specimens were permea- bilized with 0.4 % Triton-X followed by rinsing and blocking with 2 % BSA solution. Primary antibodies were diluted in 2 % BSA and bound overnight at 4 °C. After rinsing, secondary antibodies dissolved in 2 % BSA were applied for two hours at room temperature. This was followed by rinsing and mounting in Vectashield mounting media with DAPI.

After PFA fixation, tissues in Matrigel were cryopreserved and then sec- tioned into 8 µm slices and captured on SuperFrost plus slides (Menzel- Gläser). Sections were post-fixated in cold acetone for 10 minutes before staining. Sections were then air-dried and blocked with 2 % BSA. This was followed by overnight incubation at 4 °C with primary antibodies. After rinsing, sections were covered with secondary antibodies and incubated for two hours at room temperature. Specimen were then rinsed and mounted in Vectashield mounting media with DAPI.

3-D culture with coated electrodes (Paper III)

A 2 % CPHS and 1 % cellulose solution was selectively applied to the PI surface of a single PI electrode on a 2 mm CI fragment (MED-EL). The cellulose acts as a glue, binding nanoparticles to the CI. Once dry, a 0.2 µL drop of neurotrophins was applied to the CPHS surface.

Protocol for testing the coated CI surfaces is based on the protocol from paper II. Explants were encapsulated in Matrigel as described in Primary

(28)

Cultures and a coated electrode was then placed in close proximity to the explant (< 700 µm), which was cultured for up to two weeks.

Cultures on diamond substrates (Paper IV)

Each NCD surface had four different inter pillar spaces, 4, 5, 9 and 14 µm.

Human and murine material collected as described in Primary cultures were placed on uncoated, washed, and dried, diamond surfaces and a small drop of “Complete media” was applied, just covering the surface of the diamond and the explant. After 24 – 48 hours when the explant was fully attached, 2 mL media was added, filling the Petri dish. As the diamond was opaque it was not possible to follow and assess outgrowth during the culture period;

the only indication of health was whether the explant anchored. After 14 days the explants were fixated and stained.

ReNCells were maintained as described. After passaging and counting, the cells were seeded onto three different substrates: either uncoated silicon, silicon coated with 20 µg/mL laminin, or the NCD diamond substrate.

LigandTracer (Paper III)

To investigate the uptake and release of GDNF in the CPHS a ¹²⁵I labeled GDNF was used. The ¹²⁵I labeling was done using a chloramine-T reaction as previously described (Hunter & Greenwood, 1962). All in vitro slow re- lease characteristics were measured using the LigandTracer Grey Instrument (Ridgeview), detecting gamma radiation (Figure 4).

Duplicate nanoparticle-containing gels were placed inside near the rim of a 10 cm dish along with pure drop of Matrigel.

After background measurements in 3 mL of buffer, 200 µL of 1 µg/mL ¹²⁵I- labeled GDNF was added and the uptake was measured once per minute for 23 hours. Background measurements were made on untreated plastic.

For the retention experiments the ¹²⁵I-buffer was first exchanged to fresh buffer and retention was measured for seven hours at room temperature. The buffer was then exchanged again and replaced with 5 mL of fresh buffer.

Retention rate measurements then continued at 37 °C.

Data was analyzed using the TraceDrawer software.

Figure 4. Illustration of the Ligand- Tracer experimental setup.

(29)

Results

Paper I

The hNPC were investigated after 0 – 14 days of differentiation. We found that cells stained positive for nerve markers that are also found in the inner ear spiral ganglion, such as BDNF/NT-3 growth factor receptor (TrkB) and Neuron-specific class III beta-tubulin (Tuj1) (Figure 5, Table 3).

Figure 5. Immunohistochemically stained hNPC after 14 days of differentiation against A) Glutamate receptor 1 (green). B) Tropomyosin receptor kinase B (green).

C) Neurogenin 1. D) Tuj1 (red) and oligodendrocyte marker Olig1 (green). E) Tuj1 (green). F) Calcium binding protein S-100 (green). G) Phase contrast image of F.

GlutR1

DAPI TrkB

DAPI

Tuj 1 Olig1 DAPI Neurog1

DAPI

A B

C D

S-100

F DAPI

G

E ^{Tuj 1}_DAPI

100 µm 100 µm

(30)

Table 3. List of antibodies used in the study. The table also shows if cells stained positive or negative for each antibody and at what stage of differentiation staining was performed. Scoring was based on intensity and background, +++ = excellent, ++ = strong, + = positive, (+) = faint, - = negative.

Antigen Source Dilution Day 0 Pos/neg

Day 7 Pos/neg

Day 14 Pos/neg Tubulin (Tuj 1) Rabbit 1-300 + +++ +++

Tubulin (Tuj 1) Mouse 1-300 (+) +++ +++

Olig1 Rabbit 1-600 ++ ++ ++

Olig1 Mouse 1-40 ++

Brn3a Mouse 1-400 + + +

Brn3a Rabbit 1-50 +

GATA-3 Goat 2-125 - - -

GATA-3 Mouse 1-50 -

NTRK2 (TrkB) Goat 1-50 +

NTRK2 (TrkB) Mouse 1-50 (+)/- + + Neurogenin 1 Rabbit 1-50 + ++ ++

Calreg Mouse 1-50 - - -

S100 Rabbit 1-100 +

GFAP Rabbit 1-500 -

GFAP Mouse 1-500 (+)

Cx30 Rabbit 1-50 +

Glutamatreceptor Rabbit 1-100 ++

Neurofilament 160 Mouse 1-100 ++

The differentiation process was followed through TLVM analysis, which revealed the morphological stages of cell differentiation the cells underwent.

At the start, cells are clustered and fairly uniform in shape and size. Only minor morphological changes occur until day five, when colonies start con- necting to each other through thin axon-like structures. This becomes more visible at day seven when the cytoplasm also appears tighter around the nuclei and colonies are less densely packed. At day nine, morphologies vary and include nerve-like cells. At 14 days few cells remain and clusters of heterogeneous cells are connected by long axon-like structures.

(31)

Paper II

We found that human VG could be cultured as explants in Matrigel and mi- gration was followed by TLVM. The culture protocol also proved successful in promoting 3-D axonal outgrowth from murine and guinea pig explants as well as differentiated hNPC.

Figure 6. Cultured human VG. A) Phase contrast image of VG neurites extending into the 3-D matrix. B) Cryosectioned VG explant showing large neural cell bodies positive for Tuj1 (red) and TrkB (green). Scale bars 100 µm.

Various cell types migrated into the surrounding gel starting within 24 hours.

Using TLVM, it was possible to follow the long axon-like structures migrat- ing out into the gel, with growth cones finding their way through gel (Figure 6 A). Cryosectioning human explants revealed internal structures and large Tuj1 positive cell bodies could be identified (Figure 6 B). This was also done for mouse spiral ganglion where type I SG neurons were found.

Guinea pig SG neurons could also be visualized on the border between gel and explant (Figure 7 A). In rodent cultures neural outgrowth was ob- scured by cells aligning along the axonal processes (Figure 7 B).

Figure 7. Cultured guinea pig SG. A) Phase contrast image merged with immuno- fluorescent image showing Tuj1 positive neural cell bodies bordering the explant and sending extensions out into the surrounding gel (*). B) Image from TLVM.

Supporting cells align with and migrate along the nerve-like extensions. Cell divi- sion can also be observed (arrow). Scale bars 100 µm.

(32)

Paper III

CPHS nanoparticles were successfully anchored onto the CI electrodes using cellulose. The extent of the coating could be visualized by scanning electron microscopy.

When encapsulated in a gel, the GDNF uptake and release properties of the CPHS could be clearly distinguished from both the empty gel and background using LigandTracer. Results showed that the CPHS had a large load- ing capacity at room temperature and initially a rapid release rate at room temperature. The uptake and release was fitted to a Langmuir binding model and the theoretical association (k_a), and dissociation (k_d) could be calculated (Table 4).

Table 4. Summary of uptake and release experiments showing calculated maximum capacity and actual uptake as well as calculated association and dissociation con- stants.

Uptake 23 h (ng)

Maximum capacity (ng)

ka (1/(M*s)) kd (1/s) Affinity KD (nM) Average 10.4 20.1 6.3*10³ 2.6*10^-5 4.5

Stdev 2.1 6.7 1.7*10³ 4.3*10^-6 1.9

When continued at 37 °C, the release was biphasic. Initially, the release rate was high, but after 4 – 8 hours it leveled out and stabilized at a lower rate (Figure 9 A-B). The uptake and release patterns of Matrigel and cellulose indicated that within 1 hour after GDNF addition or removal they had equil- ibrated with the surrounding buffer.

When the CPHS were loaded with GDNF and placed in 3-D cultures, neurites of both human and murine origin showed affinity for the coated regions and Tuj1-positive extensions could be seen reaching for these areas. 3-D confocal scans indicated that these extensions interacted with the coated surfaces (Figure 8 C).

Figure 8. In vitro 3-D cultures with CPHS coated CI. GDNF was loaded into CPHS and selectively coated on the metal electrode. (A) The CPHS were barely visible in stereo microscope. (B-C) Tuj1-positive extensions (green) sprouting from human VG were attracted to the GDNF coated electrode 14 days after co-culture. Scale bars 100 µm.

(33)

Figure 9. GDNF uptake and release curves from LigandTracer experiments. (A) A representative curve of the uptake and release of GDNF at room temperature. (B) The continued release was monitored after the dish was moved to 37 °C. The data was normalized, and the curves show the release properties of both the measuring points at 37 °C during 16 hours for the three duplicate experiments (n=6).

A nuclear staining showed that cells growing out from a human VG explant in the 3-D gel also had affinity for the neurotrophin-loaded coating and pref- erentially followed the coated surface (Figure 10 A, B, D & E) rather than distributing randomly in the gel. Empty CPHS did not trigger such organiza- tion (Figure 10 A-F).

Figure 10. A 3-D reconstruction of Z-stacked images visualized the distribution of cells in the gel surrounding the coated electrodes. Cells were stained with DAPI (green). (A, D) The control-coated surface did not attract the growth of cells onto the electrode and the cells grew randomly near the implant surface. Both BDNF- (B, E) and GDNF-coated (C, F) CI electrode fragments attracted cells onto the coated surfaces. The co-cultures were observed vertically (A, B and C) and horizontally (D, E and F) in the gel. Metal wires are also visible as they reflect light.

A B

D E

C

F

(34)

Paper IV

NCD surfaces were successfully produced by the Department of Engineering Sciences, Uppsala University, as described in paper IV. The pillars were interspaced at 4, 5, 9 and 14 µm with a thin diamond head. Primary neurons of both human and murine origin were capable of attaching to the NCD surface without extracellular matrix coating and following the patterned diamond structures.

Immunostaining revealed Tuj1 positive extensions growing in an orderly fashion (Figure 11 A, C). When studied with 3-D confocal microscopy, the neurons appear to stay at the top of the pillars when inter-pillar distance was below 9 µm, while nuclei from non-neural cells were located between the pillars (Figure 11 B). At the 14 µm spacing neural outgrowth still followed the pillars but was instead located on the silicon surface between the pillars.

Figure 11. Mouse neurons showed high affinity for NCD pillared surface. (A–C) Confocal laser microscopy showing murine SG outgrowth (Tuj1, green) on the NCD pillars with a spacing of 4 µm. (B) 3-D rendering of Z-stacked images revealing axon growth on the top of the NCD pillars, with the cell nuclei in between. (A, C) Most sprouting neurites remained on the structured NCD surface. Dotted line indicates the border between the structured and flat NCD surfaces. Scale bars 100 µm.

VG cultures showed that human tissue also attached and projected Tuj1 posi- tive extensions along the surface (Figure 12 A-C). Patch-clamping of cells from the VG revealed that neurons were alive but not spontaneously firing (Figure 12 D).

Neurite length was assessed to find the most nerve-stimulating inter-pillar distance. In murine cultures the longest axons were observed at 4 and 14 µm.

In the human cultures only the 4 and 9 µm spacing were studied, with no discernible difference (Figure 13 A-B).

(35)

The ReNcell cultures showed that stem cells could attach to the uncoated patterned NCD surface equally or better than to a laminin-coated surface.

The optimal distance for ReNcell adhesion was 9 µm.

Figure 12. Human neurites growing on the NCD pillars expressed Tuj1 but not par- valbumin. (A) Immunofluorescence staining with Tuj1 (green) show regenerating human VG neurite growth following the NCD pillars with 4 µm spacing. (B) 3-D reconstruction of the Z-stacks revealed Tuj1 and parvalbumin (red) co-expressed in sprouted neurites. (C) The expression of parvalbumin (red) was weaker in neurites reaching the NCD surface. (D) The neuron was held at -60 mV by constant current injection. In order to evoke action potentials, a series of increasing depolarizing currents (from 50 to 300 pA in 50 pA increments, 500 ms) were injected.

Figure 13. Longest axons were observed on NCD pillared surface with 4 and 14 µm spacings. (A) Axonal length was measured using the simple neurite tracer plugin in ImageJ. Purple indicates measured axons, and the longest axon is shown in green.

(B) Diagram summarizing data of the longest axons measured from each explant.

NCD pillars with a spacing of 4 and 14 µm show a tendency towards longer axons;

however, the difference was not statistically significant. Human VG explants showed no difference in axonal length between the compared surfaces. n = 3.

Strategies in Cochlear Nerve Regeneration, Guidance and Protection: Prospects for Future Cochlear Implants